The Origin of Ferroelectricity in H f x Zr 1-x O 2 : A C om putational
Investigation and a Surface Energy M odel
R. Materlik1), C. Künneth1) and A. Kersch 1,a)
1
Munich University of Applied Sciences, Department of Applied Sciences and Mechatronics,
Lothstr. 34, 80335 Munich, Germany
_____________________________
a)
Author to whom correspondence should be addressed. Electronic mail: [email protected].
The structural, thermal, and dielectric properties of the ferroelectric phase of HfO 2, ZrO2 and
Hf0.5Zr0.5O2 (HZO) are investigated with carefully validated density functional computations.
We find, that the free bulk energy of the ferroelectric orthorhombic Pca2 1 phase is unfavorable
compared to the monoclinic P21/c and the orthorhombic Pbca phase for all investigated
stoichiometries in the HfχZr1-χO2 system. To explain the existence of the ferroelectric phase in
nanoscale thin films we explore the Gibbs / Helmholtz free energies as a function of stress and
film strain and find them unlikely to become minimal in HZO films for technological relevant
conditions. To assess the contribution of surface energy to the phase stability we parameterize
a model, interpolating between existing data, and find the Helmholtz free energy of
ferroelectric grains minimal for a range of size and stoichiometry. From the model we predict
undoped HfO2 to be ferroelectric for a grain size of about 4 nm and epitaxial HZO below 5 nm.
Furthermore we calculate the strength of an applied electric field necessary to cause the
1
antiferroelectric phase transformation in ZrO2 from the P42/nmc phase as 1 MV/cm in
agreement with experimental data, explaining the mechanism of field induced phase
transformation.
2
I. IN TR OD U C TION
The recent discovery of ferroelectricity in HfO2 and ZrO2 based high-k materials1,2 surprised
since this material class was extensively studied for decades3,4. Similar to the classical
perovskites, the origin of the ferroelectricity is most likely a non-centrosymmetric polar crystal
phase which is stable under certain conditions. Kisi5 found and structurally analyzed the polar
orthorhombic Pca21 phase (f-phase) more than 2 decades ago in ZrO2 under asymmetric stress
conditions. In this time theoretical calculations were performed 6 showing such a phase to be
stable. Neither the dielectric properties were studied experimentally nor examined by
calculations. The discovery of the ferroelectric properties arised from the development of
nanoscale thin films7 for use in memory technology, where crystalline doped films of high
quality were developed. After the observation of a remanent polarization and hysteresis in 10
nm thick films of 3 % Si doped HfO2 films1 in a TiN-HfO2-TiN stack, the effect was
additionaly found in Al- and Y-doped films8,9 grown on TiN as well. More suitable dopants
have been discovered since10. Furthermore, ferroelectricity was found in the 9 nm mixed oxide
thin film Hf0.5Zr0.5O2 (HZO)2 deposited on TiN electrodes, although neither the pure HfO2 nor
the pure ZrO2 film exhibit ferroelectricity. The existence of a true ferroelectric effect has been
confirmed by various experiments such as P-V and C-V measurements
extrapolated up to 10 years
measurements
12
8
as well as retention
and the occurrence of orthorhombic peaks in XRD
13
. ZrO2 was also studied extensively in the context of functional thin films since
its tetragonally stabilized form has a high dielectric constant. Müller11 showed, that ZrO2
3
behaves antiferroelectric: above a critical field in the order of 1 MV/cm, a temperature
dependent field induced phase transformation to the f-phase occurs. A recent review of the
current status of HfO2, ZrO2 and HZO based ferroelectric films can be found in10.
Although the ferroelectricity in binary oxides is similar to perovskites, there are some
differences which make these materials highly attractive for technological applications: the
binary oxides do not suffer from a dead layer effect which makes perovskites ineffective for thin
film technology. The mid range dielectric constant8 of the binary oxides allows switching at
moderate voltage, although the necessary field strength required for polarization reversal is
much higher than for perovskites14. The large field strength without breakdown is possible due
to the large band gap of 5.6 eV compared to 3-5 eV in perovskites15.
Despite intense research in the last years to collect phenomenological knowledge and explore
the fundamental effects, an open issue remains which is of some relevance for the fundamental
understanding as well as for the technological applications: what is the exact reason for the
stability of the ferroelectric phase in the various conditions where it has been found? As a fact,
the bulk materials have not been stabilized in the ferroelectric f-phase so far, whereas the
possible influencing factors on phase stability like temperature, doping and other defects, stress
or strain, and surface or interface energy are known. It is under intensive investigation what
specific set of circumstances is responsible for the existence of the experimental observed
ferroelectric thin films.
4
The effect of doping to promote cubic Fm-3m (c-phase) or tetragonal P42/nmc (t-phase)
phases is well known from partially and fully stabilized ceramics. The ferroelectric f-phase was
originally discovered in Si doped HfO21. So far all subsequently produced ferroelectric films
were doped as well, with the exception of HZO films. However, to completely explain f-phase
stability there is still a need for an additional mechanism since the ferroelectric properties
typically disappear in thick films and bulk materials16. In this paper, we want to focus on the
binary mixed oxide Hf0.5Zr0.5O2 to minimize computational load otherwise necessary for
calculations with dopants.
Since the discovery of transformation toughening by Garvie17, the importance of stress effects
causing the transition of stabilized tetragonal into the monoclinic P21/c phase (m-phase) is
well known in ZrO2. Asymmetric stress effects have also been discussed as an explanation for
the appearance of ferroelectric ZrO 2 grains in cubic stabilized ZrO2
18
. Furthermore, stress
effects in thin Hf0.5Zr0.5O2 films have been used as a possible explanation of how a cap
electrode affects the ferroelectric phase by Müller11,19 as well as film thickness causing film
stress variation and appearance of a ferroelectric phase by Park16.
The effect of grain size on phase stability is well known from the work of Garvie 17 and has an
important application in ZrO2 thin films contained in the high-k dielectric of DRAM
capacitors. These ZrO2 films with grain size below 30 nm are typically in the t-phase and the
involved surface energy relative to the m-phase was experimentally measured by Pitcher et al.
and other groups20,21 and calculated by Christensen and Carter22. The same effect is known for
5
HfO2 which undergoes a phase transformation at a critical particle size of about 4 nm, but the
surface energy has only been measured for the m-phase23. Müller2 observed the size effect in
ferroelectric Hf0.5Zr0.5O2 thin films by identifying the thinner 7.5 nm film as less ferroelectric
and more antiferroelectric than the thicker 9 nm film. This investigation was deepened in the
systematic work by Park24 who found a loss of ferroelectricity for films above 20 nm.
In this paper, we investigate the stability of the f-phase with density functional calculations
(DFT). Sec. II. describes the computational setup. In Sec. III A-C the computed structures,
total and free energies for various polymorphs are presented and compared with literature
data. The phonon modes and the infrared (IR) spectra are computed as well. Since absolute
values for total energies in DFT calculations are known to depend on the exchange-correlation
functional and pseudopotential on the level of a few meV/f.u. (f.u.=formula unit MO 2, M=Hf,
Zr), we validate our choice by comparison with available data. In Subsec. D the effects of
stress and strain together with elastic properties are included, and possible scenarios for a
stress or strain stabilization are discussed. In Subsec. E the effects of surface energy are
included in form of a model. The model contains free parameters in the form of surface
energies based on the limited amount of experimental and calculated data available so far. The
results are discussed in light of a variety of experimental results. Subsec. F contains the
computed effect of an electric field on the phase stability as well as the dielectric and
ferroelectric properties. The conclusions are presented in Sec. IV.
6
II. C OM P U TA TION A L M ETH OD S
All calculations are performed using the ABINIT package25,26. The exchange-correlation
energy
is
computed
with
a
local-density
approximation
using
the
Perdew-Wang
parameterization. For Hf we use a self made norm-conserving Troullier-Martin pseudopotential
generated with the fhi98 generator27 to achieve good values for the lattice constants. For Zr
and O we use well calibrated RRKJ norm-conserving pseudopotentials from the Opium project
31
. The levels Hf(5s,5p,5d,6s), Zr(4s,4p,4d,5s), and O(2s,2p) are treated as valence states with
the following atomic valence configurations for the reference state: Hf(5s 25p65d26s2),
Zr(4s24p64d05s0), and O(2s22p4). For Hf, core radii of 1.3 a0, 1.8 a0, 2.7 a0 (Bohr radii) are
chosen to describe angular waves from s to d. For Zr the cutoff radii were 1.58 a 0, 1.73 a0 and
1.79 a0 whereas for the O pseudopotential, a cutoff radius of 1.50 a0 for both s and p waves is
applied. We also adopted a separable form for the pseudopotentials treating the following
angular momentum waves as local: s for Hf, f for Zr, and p for O. For Hf, a nonlinear core
correction of radius 1.05 a0 was used28.
The Brillouin zone is sampled using the Monkhorst-Pack
29
scheme with a 6x6x6 k-point
grid for all 12 atomic phases and a 3x6x6 for the 24 atomic orthorhombic Pbca phase (ophase). The cutoff energy Ecut of the plane-wave expansion was fixed at 30 Ha. For the
calculation of the phonon modes a 2x2x2 q-point grid was calculated and interpolated to a
30x30x30 grid
30
. Integration of the phonon density of states yields the free energy. All
parameters were carefully tested for convergence. The acoustic sum rule was imposed.
7
HZO cells are constructed by replacing 50 % of Zr atoms in ZrO2 with Hf. There are two
inequivalent possibilities to build such a cell, but the difference in total energy and other
properties between both is negligible small. To save CPU time we choose one (see SI32, TABLE
SI) for our calculations.
Berry phase calculations were performed under an incrementally increasing electric field to
study the effect of an external electrical field on m-, t- and f-phase cells. Structural relaxations
are performed after each increase of 1 MV/cm in electric field strength.
III. R ESU LTS
A . Structural P roperties
HfO2 and ZrO2 are structurally and chemically similar
33
and can adopt a variety of
crystalline phases. Increasing the temperature, the monoclinic phase (m-phase, No. 14, space
group (SG): P21/c) transforms (between 1270 K and 1370 K for ZrO23 and at about 2073 K for
HfO2 3,34) into a tetragonal phase (t-phase, No. 137, SG: P42/nmc) and further (at about 2650
K for ZrO2 3 and about 2900 K for HfO23) into a cubic phase (c-phase, No. 225, SG: Fm3m). In
addition, various orthorhombic phases exist: an orthorhombic I phase (o-phase, No. 61, SG:
Pbca), an orthorhombic II phase (oII-phase, No. 62, SG: Pnma) and a polar orthorhombic III
phase (f-phase, No. 29, SG: Pca21). The phase transformation from the m- to the o-phase can
be observed at a compressive pressure of about 4-12 GPa in bulk for HfO235,36 and ZrO237,38.
The oII-phase occurs at a very high compressive pressure above 20 GPa for both HfO 236 and
8
ZrO238. We do not include this high pressure phase in our study, since such high pressures are
irrelevant for most thin film applications. The transformation to the polar f-phase can be
observed under asymmetric stress conditions18 which will be discussed in section C. Although
the unit cell of the c-phase has only 3 and the t-phase only 6 atoms, we represent all the
structures except the o-phase in 12 atom cells for a better visualization of the structural
similarity. The Wyckoff coordinates are shown in TABLE I and the structures in FIG. 139. The
polarization of the f-phase can be visualized in the form of the displacement of the four O1
atoms in the z-direction after polarization reversal by mirror reflection, whereas the metal and
O2 atoms are hardly displaced. When the motion of an atom i in direction j is rij relative to
the mirror symmetric, polarization neutral configuration, the polarization Pj of the structure is
Pj 
e 12 *
  Z ij  rij
V i 1
(1)
with cell volume V, Born charge tensor Z*ij and unit charge e. Upon a point reflection the
polarization changes as well, but the resulting structure is unequal to the mirror reflected
structure. The reason is that the f-phase appears to occur in two different forms with opposite
chirality. We are unaware of any physical reason to discriminate in favor of any particular
chiral form as both structures have the same total energies and hence decided for the mirror
reflected structure shown in FIG. 1 based upon better visualization of the polarization reversal.
Furthermore, the 24 atomic unit cell of the o-phase can be constructed by gluing the two
illustrated unit cells of the f-phase with opposite polarization and chirality alongside the xy-
9
plane together and relaxing the structure. Therefore the o-phase has one additional symmetry
compared to the f-phase.
39
FIG. 1. Crystal phases and experimental lattice constants in Å for ZrO2. Zr atoms are represented in green
(big), O atoms in red (small). In the f-phases the O atoms mainly responsible for the polarization are highlighted
golden. To illustrate the relationship between the f- and o-phase, the corresponding O atoms are highlighted in
the o-phase as well. The polarization axis of the two f-phase cells is marked by black arrows (P).
The similarity between structures can be better understood by comparing the relative
coordinates of the Wyckoff positions in an appropriate representation (choice of origin). The
chosen representation allows the construction of an initial reaction path for the calculation of
the phase transformation to determine the transition state. The relative coordinates and cell
parameters of the calculated structures are shown in TABLE I. The representations of the
structures were checked with FINDSYM
40
.
10
41
TABLE I. Calculated lattice constants a, b, and c for a unit cell in Å. a’, b’, and c’ are the lattice constants of
3
a 12 atomic cell with less than 12 atoms per unit cell. The cell volume V is in Å , and the total energy difference
*
U in meV/f.u. The lattice constants are aligned with the direction x, y, z, and z (given in Miller vector notation)
*
as defined in FIG. 1.  is the angle of the m-phase between x and z . The Wyckoff positions of the Metals (M)
and Oxygen atoms (O1 and O2) are given in relative coordinates. The arrows indicate the polarization direction
of the f-phase.
c-ZrO 2
U = 96
c-H fO 2
V = 129.2
x= [100] y= [010] z= [001]
U = 112
V = 128.4
x= [100] y= [010] z= [001]
a’= 5.05 b’= 5.05 c’= 5.05 a’= 5.04 b’= 5.04 c’= 5.04
M b
0.250
0.000
0.250
0.250
0.000
0.250
O1 c
0.000
0.250
0.000
0.000
0.250
0.000
O2 c
0.500
0.250
0.500
0.500
0.250
0.500
t-ZrO 2
U = 49
t-H fO 2
V = 13 1.8
U = 92
V = 131.1
x= [110] y= [001] z= [110] x= [110] y= [001] z= [110]
a’= 5.06 c= 5.15 b’= 5.06 a’= 5.05 c= 5.14 b’= 5.05
M b
0.250
0.000
0.250
0.250
0.000
0.250
O1 d
0.000
0.294
0.000
0.000
0.293
0.000
O2 d
0.500
0.206
0.500
0.500
0.207
0.500
m -ZrO 2
U = 0
m -H fO 2
= 80.5° U = 0
V = 13 8.5
V = 138.4
= 80.3 °
*
x= [100] y= [010] z = [001] x= [100] y= [010] z * = [001]
a= 5.11
b= 5.22 c= 5.27
a= 5.12
b= 5.18 c= 5.29
M 4e
0.278
0.043
0.290
0.276
0.042
0.292
O1 4e
0.074
0.334
0.165
0.068
0.331
0.153
O2 4e
0.553
0.258
0.481
0.550
0.257
0.480
f-ZrO 2 ↓
U = 3 7
f-H fO 2 ↓
V = 13 3 .8
U = 62
V = 133.5
x= [010] y= [100] z= [001]
x= [010] y= [100] z= [001]
b= 5.04
b= 5.04
a= 5.25 c= 5.06
a= 5.23 c= 5.06
M 4a
0.268
0.032
¼+ 0.002
0.267
0.032
¼+ 0.007
O1 4a
0.073
0.369
¼-0.13 3
0.068
0.389
¼-0.138
O2 4a
0.539
0.266
½+ 0.006
0.537
0.267
½+ 0.008
f-ZrO 2
U = 3 7
↑
f-H fO 2 ↑
V = 13 3 .8
U = 62
V = 133.5
x= [010] y= [100] z= [001]
x= [010] y= [100] z= [001]
b= 5.04
a= 5.04
a= 5.25 c= 5.06
b= 5.23 c= 5.06
M 4a
0.268
0.032
¼-0.002
0.267
0.032
¼-0.007
O1 4a
0.073
0.369
¼+ 0.13 3
0.068
0.389
¼+ 0.138
11
O2 4a
0.539
0.266
½-0.006
0.537
o-ZrO 2
U = 26
0.267
½-0.008
o-H fO 2
U = 24
V = 13 8.7
x= [100] y= [010] z= [001]
V = 138.4
x= [100] y= [010] z= [001]
a/2= 5.07 b= 5.20 c= 5.27 a/2= 5.07 b= 5.15 c= 5.29
M 8c 0.278/2
0.042
0.340
0.276/2
0.041
0.343
O1 8c
0.069/2
0.338
0.174
0.064/2
0.325
0.160
O2 8c
0.552/2
0.249
0.589
0.550/2
0.248
0.589
A martensitic phase transformation between crystal structures requires a specific spatial
orientation due to constraints by transition path and barrier. The orientation between the tand m-phase during temperature driven-phase transformations has been investigated
experimentally and theoretically
42,43
. It was found that the [010]m || [001]t directions and the
(100)m || (110)t planes coincide. In addition the spatial orientation between the t- and f-phase
has been investigated by Kisi
4
who found a similar coincidence of [010]f || [001]t and (100)f ||
(110)t. The spatial orientation between the f- and o-phase is derived from the representation in
FIG. 1 and implies a spatial relation between the o- and m-phase via the f- and t-phase.
In the following discussion, we will focus on ZrO2 to further exemplify the relation between
the various phases (Values from TABLE I). The m→t transition involves a volume change
from 138.5 Å3 to 131.8 Å3 and the change of c from 5.27 Å to 5.06 Å. Besides the difference of
the lattice angle , the main distinction between the m- and the f-phase is a polarization
dependent displacement of the O1 atom in z-direction. The o-phase is closely related to the fphase because the 24 atomic cell is composed of two oppositely polarized 12 atomic f-phase
cells. Nonetheless the cell volume and lattice constants differ: the volume changes from 138.7
12
Å3 to 133.8 Å3 and the length of the polarization axis c from 5.27 Å to 5.06 Å. The
experimental results show only a small difference in volume and length between these two
phases. A transformation from the m-phase to the f-phase involves a volume change from 138.5
Å3 to 133.8 Å3 and a reduced length of the polarization direction c from 5.27Å to 5.06Å. The
transformation from the t-phase to the f-phase involves only a volume change from 131.8 Å3 to
133.8 Å3 and the polarization direction c has almost no change. At this point, a transformation
to the f-phase seems to be easiest from the t-phase although both phases have no close
symmetry relationship. Comparing Hf and Zr, all the HfO 2 cells are slightly smaller than the
ZrO2 cells consistent with the lanthanide contraction argument
33
.
It is known that the accuracy of the calculated results depend on the chosen density
functional
and
pseudo-potential.
While
lattice
constants
from
generalized
gradient
approximation (GGA) calculations are typically too large, the results from local density
approximation (LDA) calculations are typically too small (about 1 %). However, the
discrepancy for the LDA decreases, when the expansion effect of zero point motion is included
44
. Furthermore, the calculated results depend on the chosen basis set and the number of k-
points which were not always well documented in previous studies. In TABLE II, III, and IV
we compare our calculated results for lattice constants of HfO2, HZO, and ZrO2 respectively,
with calculated values and with experimental data found in literature about the investigated
phases. For HZO, only our own calculated values are available as well as experimental data
from Müller et al.
11
. In all cases, except for the c-phases, our structural values have proved to
13
be better than 1 % compared to experimental data without taking expansion effects from zero
point motions into account.
14
TABLE II. Comparison lattice constants a, b, and c in Å and total energy difference ΔU in meV/f.u. for HfO2
Structure
U a
m-HfO2
t-HfO2
b
f-HfO2
o-HfO2c
c-HfO2
b
method
reference
/
V
3
[Å ]
13 7.
9
13 5.8
0
137.1
26.6
5.11
5.16
5.28
LDA
0
136.1
26.4
5.09
5.16
5.26
LDA
46
0
125.3
24.8
4.95
5.06
5.08
LDA
47
0
130.9
-
-
-
-
LDA
48
0
139.0
26.8
5.13
5.19
5.30
GGA
48
0
139.8
26.9
5.14
5.20
5.31
GGA
49
0
138.2
26.6
5.12
5.19
5.28
PAW
U a
/
V
13 3 .1
Axz
25.6
a’
5.06
c’
5.20
b’
-
exp.
45
/
13 3 .1
25.6
5.06
5.20
-
exp.
50
92
129.5
25.3
5.03
5.12
-
LDA
[meV]
/
Axz
2
[Å ]
a
[Å]
b
[Å]
c
[Å]
26.7
5.12
5.17
5.29
exp.
45
26.4
5.07
5.14
5.29
exp.
11
(*)
46
(*)
71
118.8
24.0
4.90
4.95
-
LDA
47
>60
135.2
25.6
5.06
5.28
-
LDA
49
99
125.1
-
-
-
-
LDA
48
156
133.1
25.6
5.06
5.20
-
GGA
138
130.3
25.3
5.03
5.15
-
PAW
U a
V
Axz
b
a
c
/
-
-
-
-
-
exp.
62
48
24
132.1
121.0
138.1
25.3
23.9
26.1
5.02
4.88
5.10
5.22
5.07
5.30
5.04
4.89
5.11
LDA
47
LDA
6
LDA
>30
134.6
25.5
5.01
5.29
5.08
GGA
63
102.8
24.1
4.90
5.11
4.92
PAW
U a
V
13 2.
6
Axz
a/2
b
c
26.2
5.01
5.06
5.23
exp.
51
LDA
(*)
/
48
46
(*)
49
46
24
138.4
26.7
5.07
5.15
5.29
60
137.8
27.0
5.08
5.11
5.31
LDA
29
126.0
-
-
-
-
LDA
65
134.1
25.4
4.92
4.96
5.16
GGA
48
>25
134.4
26.5
5.02
5.08
5.27
GGA
49
U a
V
Axz
a’
b’
c’
/
13 1.1
25.8
5.08
-
-
exp.
52
137
127.3
25.3
5.03
-
-
LDA
(*)
93
116.9
23.9
4.89
-
-
LDA
47
152
123.0
-
-
-
-
LDA
48
237
129.6
25.6
5.06
-
-
GGA
6
48
48
15
208
127.3
a
25.3
5.03
-
-
b
PAW
46
c
(*) this work, relative to m-HfO2, values for 12 atom cell, half c-axis
TABLE III Comparison lattice constants a, b, and c in Å and total energy difference ΔU in meV/f.u.for HZO.
Structure
U a
V
3
[Å ]
Axz
2
[Å ]
a
[Å]
b
[Å]
c
[Å]
Method
reference
0
137.6
26.6
5.11
5.18
5.28
LDA
(*)
U a
V
Axz
a’
c
b’
70
130.3
25.4
5.04
5.13
=a
LDA
(*)
U a
V
Axz
b
a
c
/
13 2.3
25.3
5.01
5.24
5.04
exp.
49
132.8
25.4
5.03
5.23
5.05
LDA
(*)
U a
V
Axz
a/2
b
c
25
137.1
26.7
5.06
5.14
5.27
LDA
(*)
U a
V
Axz
a’
b’
c’
119
127.3
25.3
5.03
=a
=a
LDA
(*)
[meV]
m-HZO
t-HZO
b
f-HZO
o-HZO
c
c-HZO
b
a
b
11
c
(*) this work, relative to m-HZO, values for 12 atom cell, half c-axis
16
TABLE IV. Comparison lattice constants a, b, and c in Å and total energy difference ΔU in meV/f.u.for ZrO2.
Structure
U a
[meV]
m-ZrO2
t-ZrO2
b
f-ZrO2
o-ZrO2
c
c-ZrO2
b
/
V
3
[Å ]
Axz
2
[Å ]
a
[Å]
b
[Å]
c
[Å]
Method
reference
140.3
27.0
5.15
5.20
5.32
exp.
4
(*)
0
138.2
26.6
5.11
5.20
5.28
LDA
0
136.6
26.3
5.09
5.20
5.24
LDA
47
0
136.7
-
-
-
-
LDA
48
0
136.1
26.3
5.09
5.18
5.24
LDA
0
144.7
27.4
5.20
5.28
5.35
GGA
48
0
144.1
27.5
5.19
5.24
5.38
GGA
53
U a
V
Axz
a’
c
b’
/
13 4.2
25.9
5.09
5.18
=a
exp.
4
/
13 3 .4
25.8
5.08
5.17
=a
exp.
49
131.3
25.5
5.05
5.15
=a
LDA
34
130.1
25.4
5.04
5.12
=a
LDA
47
38
130.4
-
-
-
-
LDA
48
50
129.3
25.3
5.03
5.11
=a
LDA
109
137.9
26.3
5.13
5.21
=a
GGA
48
112
138.4
26.2
5.12
5.28
=a
GGA
53
U a
V
Axz
b
a
c
/
13 5.5
25.8
5.07
5.26
5.08
exp.
5
37
133.4
25.5
5.04
5.24
5.05
LDA
34
132.1
25.3
5.02
5.22
5.04
LDA
-60
138.1
26.1
5.10
5.30
5.11
LDA
U a
V
Axz
a/2
b
c
/
13 4.9
26.5
5.04
5.09
5.26
exp.
25
138.7
26.8
5.07
5.20
5.27
LDA
-99
137.8
27.0
5.08
5.11
5.31
LDA
14
131.9
-
-
-
-
LDA
26
130.8
26.0
4.99
5.03
5.21
LDA
49
138.8
26.1
5.09
5.14
5.31
GGA
48
67
138.1
26.9
5.08
5.13
5.30
GGA
53
U a
V
Axz
a’
b’
c’
/
13 4.2
26.2
5.12
-
-
exp.
120
127.3
25.3
5.03
-
-
LDA
82
127.3
25.3
5.03
-
-
LDA
47
67
128.6
?
?
-
-
LDA
48
94
127.3
25.3
5.03
-
-
LDA
171
134.8
26.3
5.13
-
-
GGA
48
215
134.9
26.2
5.12
-
-
GGA
53
a
b
53
11
(*)
53
(*)
47
6
54
(*)
6
48
53
55
(*)
53
c
(*) this work, relative to m-ZrO2, values for 12 atom cell, half c-axis
17
For the calculation of the force required for a stress induced phase transformation, it is
important that the spatial orientations of the cells match together for the phase
transformation, and the calculated cell parameters and volumes agree with experimental data
at least in their relative values to each other. Otherwise values for the necessary pressure or
stress will be inaccurate. For the plane strain, the cell areas in the compressed or expanded
planes are relevant. The comparison of our calculated volume ratios to the experimental
volume ratios are summarized in the supplementary material (SI32, TABLE SII) shows good
agreement.
B . Total Energy
The structural relaxation provides values for the specific total energy per f.u. for each phase.
TABLE I contains the calculated total energy differences U relative to the m-phase. These
values constitute the foundation of the calculation of the phase stability since the relevant
criteria like the Helmholtz free energy and the Gibbs energy are calculated from this difference
by adding a temperature or pressure contribution. The total energy values are only indirectly
accessible in experiments and their consistency can only be concluded from the correctness of
the final result. LDA results repeatedly in smaller energy differences than GGA calculations.
These differences have been systematically studied by Jaffe et al.
and by Fadda et al.
53
48
for both HfO2 and ZrO2,
for ZrO2 both using ultrasoft pseudopotentials. Considering the total
energy, Jaffe et al. concluded that GGA results are needed since with LDA the phase
transformation pressure between the m- and o-phase turned out to be negative. Fadda et al.
18
compared the calculated results with plenty of data sources and found LDA as a better overall
description but found too small values for the transition pressure between m- and o-phase as
well. A comparison of their results for U in TABLE IV shows that Jaffe et al. obtained
systematically smaller values for all phases and compounds than Fadda et al. Also, their with
LDA calculated volume ratios of the structures are not in good agreement with experimental
results. Therefore the comparison shows the dependence of the correctness of the conclusion on
the quality of the pseudopotentials.
The results for U show consistently positive values for all authors for the stability of f-HfO2.
However, Lowther6 revealed large negative values for f-ZrO2 and even larger negative values
for the o- ZrO2, implying stability of o-phase. With this single exception, all calculated results
lead to same order of increasing total energy values: U(m)<U(o)<U(f)<U(t)<U(c).
C . H elm holtz Free Energy
To assess the phase stability at a finite temperature T and entropy S, we calculated the
Helmholtz free energy F as
F = U - TS
(2)
from the phonon contribution by integrating over the phonon density of state. Fadda
Zhao
56
53
and
have argued, that LDA gives better results for phonon mode frequencies than GGA.
The quality of our LDA and the choice of pseudopotentials can be estimated by comparison of
the calculated IR-mode frequencies with an experiential IR-absorption spectrum (see FIG. 2 a19
c). With a slight underestimation of the computed frequencies the agreement with
experimental data is generally good. The similarity of the structures of Hf1-χZrχO2 becomes
visible in the IR-absorption spectrum for χ = 0, 0.5 and 1 (see SI
32
, FIG. S1).
The IR-absorption spectra as well as the Raman spectra are fingerprints to identify different
phases. The difficulty to distinguish the orthorhombic f- and o-phase from the XRD spectrum
has been discussed by Howard
57
. In contrast, according to the ab initio results the t-, f- and
the o-phase might be more easily distinguished using IR- or Raman-spectra instead of XRD
(see SI
32
, FIG. S1 a-c).
FIG. 2. (a) Calculated IR-Spectrum of (a) m-ZrO2, (b) t-ZrO2 and (c) m-HfO2 in comparison with experimental
59
.
58
The results for the Helmholtz free energy difference F(T) of the phases relative to the mphase as a function of temperature are shown in FIG. 3 a-c for HfO2, HZO, and ZrO2.
Qualitatively similar results for HfO2, with the exception of the c-phase, have been obtained by
Huan
49
, and for ZrO2 by Fadda et al.
53
with the exception of the f- and c-phase. For all
values χ in Hf1-χZrχO2 the m-phase has the lowest Helmholtz free energy until the mt
20
transition temperature (2100 K, 1750 K and 1250 K for HfO2, HZO and ZrO2 respectively
compared to 2073 K, 1700 K and 1300 K in experiments
3,34
). At a slightly higher temperature
(the tc transition temperature) the c-phase acquires the lowest Helmholtz free energy (2900
K, 2775 K and 2650 K in experiment
3
compared to 1950 K, 1550 and 1100 K in calculation
for HfO2, HZO and ZrO2 respectively). Our transition temperatures for the mt transition are
in good agreement with experimental data. However, the order of the mt and tc
transformations is reversed in our calculation, indicating either a deficiency of the chosen
pseudopotentials for the c-phase or a limited validity of the harmonic approximation for large
temperatures. All Helmholtz free energy differences are larger in HfO2 by about 20 % compared
to ZrO2.
More important for our paper is the result, that neither the f- nor the o-phase becomes favored
by a increase in temperature, although the f-phase is only about 50 meV/f.u. and the o-phase
only about 25 meV/f.u. less favorable than the m-phase. This is consistent with the results of
Huan49 for HfO2 and can be understood by an entropy argument: the t- and c-phase lose their
higher symmetry faster with increasing thermal motion than the m-, o-, and f-phase and are
therefore involved in thermally driven phase transformations. The degree of symmetry of the
m-, o-, and f-phase is the same, measured by the number of symmetry operations per atom
(=0.333), explaining their nearly temperature independent Helmholtz free energy difference.
Therefore the o-phase can not be favored in a thermally driven phase transformation.
21
FIG. 3. Calculated total energy differences U (o) and Helmholtz free energy differences F(T) relative to the mphase as a function of temperature, for (a) HfO 2 (b) HZO and (c) ZrO2.
22
D . Strain and Stress G ibbs Energy
Since it has been established that temperature does not stabilize the f-phase alone, we now
investigate stress effects for a possible stabilization mechanism. Using an isotropic pressure
constraint the phase stability is determined by the Gibbs energy G as
G = F + pV = U – TS + pV
(3)
with the pressure p and the volume V. Diamond-anvil cells have been used to determine
pressure driven phase transformations experimentally. For HfO236,60 and ZrO238,54, the phase
transformation from the m-phase to an orthorhombic phase has been found between 4-12 GPa.
After a long dispute, the phase was identified as the o-phase in both cases
4,37
. To obtain a
good value of the m→o transition pressure in the simulation the total energy difference
between the o- and m-phase, and the value of the bulk modulus must be calculated accurately
enough (48,53 and discussion in sect. B). Our calculated values for transition pressure and bulk
modulus compared to experimental data and other calculated values from the literature can be
found in SI
32
, TABLE SIII. With values between 10-15 GPa for the m→o phase
transformation we obtain a good agreement. The phase transformation pressure has been
derived from values of the Gibbs energy difference G at 300 K relative to the m-phase in
FIG. 4(a). The m-phase is stable around zero pressure. For compressive pressure above 10-15
GPa the o-phase is favored. The t-phase becomes stable for tensile pressures, but the f-phase is
never stabilized. The Gibbs energy and hence the transition pressure is modified by effects of
23
the surface energy as known from diamond-anvil experiments with 30 nm small grain material
from Ohtaka54. The model for the inclusion of the surface energy to the Gibbs / Helmholtz free
energy will be discussed in Subsection E. FIG. 4(b) illustrates calculated values for a grain size
of 30nm. The surface effects lower the phase transformation pressure.
FIG. 4. (a) Calculated Gibbs energy difference G relative to the m-phase around zero pressure and m→o
transition for compressive pressure. (b) Gibbs energy difference including the surface energy G for 30nm grains
according to our model.
The fact that the f-phase has only been observed in thin films suggests that the phase could
be stabilized under anisotropic strain conditions determined from a fixed surface area and zero
stress in the normal direction of the film. The stable phase under any strain constraint within
the three strain planes xy, xz, and yz, is the one with the lowest value of the Helmholtz free
energy61. In principle, all combinations of strain planes have to be considered while comparing
the strain energy of two phases. However, an energy crossover can only happen in spatial
orientations of cells with a rough match of lattice constants. For our calculations of the energy
crossover under strain we have chosen the spatial orientations of the 12 atomic cells as shown
24
in TABLE I. The spatial orientation between the c- and t-phase is obvious, as is the spatial
orientation between o- and f-phase. For the t-phase, the c-axis in the crystallographic [001]
direction is the largest with 5.15 Å and conforms most easily to the a-axis in [100] direction of
the f-phase of length 5.25 Å. The a- and b-directions of length 5.06 Å conform to the b- and cdirections of the f-phase of length 5.04 Å and 5.06 Å. The remaining two possibilities of a
match: transformation (a(t), c(t), b(t))→(b(f), a(f), c(f)) (choice in TABLE I) and c(f) and b(f)
inverted are obviously equivalent. Therefore a t→f transformation requires only a moderate
elongation of the c(t)-axis and no significant modification of the a(t) and b(t) axis and could be
promoted with a strain in the xz- or yz-plane. Crucial is the strain energy of the m- and ophase since these start with the lowest energy from the beginning. The relative orientation
between the m- and t-phase has been widely discussed in the literature in context of
transformation toughening. Our choice for m→t in TABLE I corresponds to the findings of42,43
(a(m), b(m), c(m))→( a(t), c(t), b(t)). An alternative orientation (a(m), b(m), c(m))→(a(t), b(t), c(t))
was favored by Luo et al.
34
while he was looking for the smallest transition barrier between
the m- and t-phase. We have calculated the total energy U for the xz- and yz-strain planes,
and all phases for the chosen orientation. The values of the lattice constants and the
corresponding cell areas A have been chosen by modifying the in plane stress. In equilibrium,
the grains should then adopt the crystal phase with the minimum Helmholtz free energy. In
FIG. 5 a-b we show the results as a function of cell area A for ZrO2. For HfO2 and HZO see
SI32, FIG. S2. In our chosen orientation, only a small window of stability occurs for the f-phase
in the yz-plane. The crossover point corresponds to a film stress of about 3 GPa in the f-phase
25
and 8 GPa in the m-phase in the yz-plane (label m). No crossover exists in the xz-plane. In the
orientation chosen by Luo et al.34 there is a large window of stability in xz-plane, with
corresponding plane stress of 9 GPa in the m-phase (label m’) and near zero stress in f-phase.
In the yz-plane both orientations coincide.
FIG. 5. (a) Calculated U for all phases with plane strain in the xz-direction. Indicated are the xz-areas from data
(o) compared to the xz-areas from the calculation (+). (b) Calculated U for all phases with plane strain in the
34
47
yz-direction. The curves m’ represent the m-phase in the orientation chosen by Luo et al. and others while m
refers to our choice in Table I. For comparison (b) contains the third possible orientation m” which stabilizes the
m-phase for all possible strain conditions.
To obtain an impartial argument we have calculated the Helmholtz free energy F for all
strain planes xy, xz, and yz, and all phases. The results can be found in the SI32, FIG. S3.
26
FIG. 6. (a) Calculated U for m-, f-, t-, and o-phases of HZO with plane strain in the 45° rotated plane. (b)
Calculated U ZrO2 with plane strain in the 45° rotated plane.
The conclusion is that a strain induced f-phase is only stable for a compressive strain in the
xz-plain when the m-phase orientation of TABLE I is suppressed for some reason (see FIG.
5a). Unfortunately, for such an orientation the polarization direction is within the thin film
plane. Therefore the grains can not be polarized by an external electric field perpendicular to
the thin film plane and an existing polarization can not be detected.
So far the strain was oriented in the x, y, and z directions only. Tipping the c-axis out of
the plane so that a measurable polarization could remain was suggested by Park16. In FIG. 6
we have calculated how an angle of 45° affects the phase stability by rotating the stress tensor
around the x-axis. No distinction between m-phase and m’-phase is necessary since both
orientations fall upon each other. A window of stability opens for the f-phase (compared to
FIG. 5) but the crossover points to the m-phase are located at a corresponding stress of 8 GPa
for HZO and 7 GPa for ZrO2. Based on the high corresponding stress values for crossover
points indicated by our calculations we conclude that the film stress is unlikely the most
27
important factor to cause a phase stabilization of the f-phase with a metastable polarization
along the direction perpendicular to the thin film plane.
The remaining question is whether the calculations are consistent with the conditions of the
first experimental observation of the f-phase in Mg-PSZ (Mg doped, partially stabilized
zirconium) by Kisi
18
. The material was described as a matrix of Mg-stabilized c-phase
containing grains in the t-phase. Upon cooling, the cubic matrix thermally contracts uniformly,
whereas the c-axis of the embedded t-phase grains contracts faster, the a-axis and b-axis
slightly slower. This results in a tensile stress condition on the t-phase grains in the c-direction
of about 3.6 GPa and a small compressive stress in the a- and b-direction of about 0.5G Pa.
FIG. 7. (a) Calculated G relative to the m-phase for all phases and all unidirectional stress directions x, y and z
for ZrO2 at 900 K. The included total energies are from the pure oxide without effect of dopant. (b) Calculated
G for ZrO2 at 0 K. The arrows indicate that the m-phase is destabilized in experiments due to Mg-doping 18.
Stable range of f-phase indicated by (o).
By using the volume V0 of the zero stress cell, the stress tensor σij, and strain tensor sij we
have calculated the anisotropic Gibbs energy as
Gi = F + V0σiisii = U – TS + V0σiisii ,
i = x, y, z
(4)
28
and the Gibbs energy differences Gi relative to the m-phase for all phases and stress
directions, keeping the stress in the orthogonal directions as zero. The results in FIG. 7 show
the pressure dependent Gibbs energy at 900 K and 0 K. Since in the experiment the Mg
doping stabilized the enclosed grains in the t-phase at room temperature and above, we assume
that the doping shifts the Gibbs energy G of the m-phase above the Gibbs energy G of the tphase. Reducing the temperature shifts the t-phase up relative to the f-phase due to the
entropy contribution TS. In FIG. 7 the f-phase is stabilized at a compressive stress. Our
calculations neither included the effects of Mg dopants nor the compressive stress in a and b
direction, which might open a larger window of stability for the f-phase. Therefore, we
conclude that the computed results are in reasonable agreement with the experimental findings
of a f-phase under unidirectional, tensile stress conditions. The major conclusion out of this
section is that the f-phase with a measurable polarization in the normal direction cannot be
stabilized with compressive film stress alone.
E. Surface Energy and Energy C rossover
It is known that surface energy effects are responsible for size driven phase transformations
on the nanoscale17. An example is the transformation of nanocrystalline ZrO2 from the m- to
the t-phase which is utilized to increase the capacitance of nowadays DRAM capacitors. The
surface energies  of m- and t-ZrO220,21 as well as m-HfO223 have been measured, leading to a
consensus at least about the range of the values between 2-4 J/m2. Surface energies  have also
been calculated for various orientations by Christensen22 and range between 1.246-2.464 J/m2
29
for m-ZrO2 and 1.239-1.694 J/m2 for t-ZrO2, but the medium surface orientation of the grains
is not known. TABLE V contains a collection of literature results. The surface energy
introduces a further dependence in the Helmholtz free energy F (or the Gibbs energy G, when
the pressure is included) and the phase stability of nanocrystals of surface area Ω is determined
from
F = F +  Ω = U – TS +  Ω
(5)
The requisite for the existence of t-ZrO2 nanocrystallites is a smaller surface energy
compared to the m-phase, which is confirmed by the data in TABLE V and consistent with a
size driven phase transformation and energy crossover below a diameter of about 24 nm62. In
the case of HfO2 the energy crossover to the t-phase has been observed below about 3 nm
63
.
The surface energy is also dependent on the termination with H or OH (anhydrous or hydrous)
or other radicals. As a consequence, an exposure of tetragonally stabilized ZrO 2 with water
could lead to a transformation back to the m-phase, an effect known as low-temperature
degradation64. Finally, the surface energies have been measured in powder whereas many
applications use nanocrystalline thin films with additional effects from interface energy. The
essential observation in the size driven m→t transformation in ZrO2 compared to HfO2 is that
in HfO2 a much larger surface area to volume ratio is required. This implies a smaller
difference between the surface energy of the m- and t-phase in HfO2 compared to ZrO2.
30
TABLE V. Surface energy  of Hf1-χZrχO2 for the competing phases from data calculation if available and from
our model (bold).

(H fO 2 )
[J/m 2 ]
 (H ZO)
2
 (ZrO 2 )
[J/m ]
[J/m 2 ]
m -phase, model
3.4
data (anhydrous) 3.70.123
data (hydrous)
2.80.123
data
calculation
-
3.2
3.0
3.450.2865
2.860.3165
6.40.220
1.246-2.46422
o-phase, model
3.3
2.9
2.5
f-phase, model
3.15
2.575
2.0
t-phase, model
3.1
2.5
1.9
data (anhydrous)
1.030.0565
data (hydrous)
1.230.0465
data
2.10.0520
calc
c-phase, model
-
-
1.239-1.69422
3.05
2.4 25
1.8
Since our question concerns the stability of the f-phase we need values for the surface
energy of all the competing phases which have neither been measured nor calculated.
Nonetheless, we need values for the composition dependence since the ferroelectric phase is
observed in HZO. We chose the missing values as model parameters. The choice has been
guided by the existing data, most importantly by the smaller difference of  (m-HfO2) and  (tHfO2) compared to  (m-ZrO2) and  (t-ZrO2). Additionally, we decrease  with increasing
crystal symmetry. The composition dependence finally is modeled from linear interpolation
 (Hf1-χZrχO2) = (1-χ)  (HfO2) + χ  (ZrO2)
(6)
After our choice of values for the surface energy as given in TABLE V we now have a
complete model to calculate the phase stability of Hf1-χZrχO2 and are able to compare with
existing thin film data. Assuming the absence of a pressure or strain constraint we determine
the stable phase from the minimum of the Helmholtz free energy including the surface
31
contribution F = U – TS +  Ω. The total energy U and the entropy contribution TS are
calculated fundamentally from DFT as above. At this point it is clear that these final results
depend on the accuracy of the total energy values and entropy contributions, justifying the
lengthy discussion above. The surface contribution Ω finally uses a phase and composition
dependent model parameter  and a geometry model for the calculation of the surface area Ω,
assuming a cylinder surface with height h from film thickness and radius r=h/2, if not given
otherwise.
We continue with a comparison of the model results with data. TABLE VI contains the
calculated results for a 9 nm thin film at 80 K, 300 K and a 900 K anneal temperature
modeling the experimental results of Müller
11
. For all temperatures the 9 nm thick film is
calculated to be monoclinic even at anneal temperatures. For smaller crystallites Shandalov63
observed a phase transformation between a size of 7.6 nm and 3.1 nm. For 3.1 nm crystallites
we calculate that f-HfO2 has the lowest Helmholtz free energy in close competition with tHfO2. Since the crystallites have been annealed we have calculated the values at 900 K and
find t-HfO2 with the lowest Helmholtz free energy. We hypothesize that such crystallites can
be transformed to f-HfO2 under the influence of an E-field. Finally, we calculate the values for
crystallites of 2 nm diameter observed by Bohra et al.
66
. At this small size we clearly find f-
HfO2 to have the lowest Helmholtz free energy. Bohra et al. identified the crystallites to be of
o-phase from an electron diffraction pattern. A possible explanation is, that for extremely small
32
crystallites a further model is required which adds the effect of size induced hydrostatic
pressure
67 68
.
The next results in TABLE VI model the 9 nm thin film HZO from Müller et al. at 80 K,
300 K and 900 K. At 80 K f-HZO is stable with an energy of 7 meV/f.u. below the competing
t-HZO. At 300 K the f-HZO is still stable but only 1 meV/f.u. lower than the competing tHZO. At anneal temperature the t- and c-phase are more favorable, and a phase
transformation to the f-phase has to take place upon cooling. The calculations for a 7 nm film
show the tendency of a thinner film to become tetragonal. The data for a 7 nm film from
Müller
2
compared to a 9 nm film show this tendency in a rudimentary antiferroelectric
hysteresis curve. This size driven transformation from t- over f- to m-HZO is shown in the
calculation and data of Park
16
who investigated a thickness series of 5.5 nm, 9.2 nm, 14 nm,
19 nm, and 24 nm.
The last group of results in TABLE VI is about ZrO2 starting again with a model for the
data from Müller
11
. The 9 nm film is paraelectric at 80 K and room temperature. The
antiferroelectric character becomes visible in an electric field driven paraelectric to ferroelectric
phase transformation. The strength of the required electric field in Müller’s data decreases
from about 1 MV/cm to about 0.5 MV/cm when lowering the temperature from 230 K to 80
K. In the calculation the film shows to be tetragonal in both cases, but the energy difference to
the f-phase lowers from 10 meV/f.u. to 5 meV/f.u. which is consistent with a decrease of the
required electric field. The last data in the table concern the crossover size from the t- to the
33
m-phase. According to the data of Yashima
62
a diameter of 24 nm is sufficient whereas in the
model a size of 36 nm is necessary to yield the m-phase to have the lowest energy.
34
TABLE VI. Comparison of F relative to the m-phase for HfO2 crystallites of different size and temperature in
comparison with data. Bold numbers point out the lowest Helmholtz free energy according to our model.
T
[K ]
h/r
[nm ]
calculated F in meV/f.u.
m
o
f
t
c
observed
phase
H fO 2
3
300
bulk
0
23
59
78
113
m-phase
80
9/4.5
0
11
30
50
83
m-phase
11
300
9/4.5
0
11
29
42
70
m-phase
11
900
9/4.5
0
8
24
16
28
m-phase
11
300
7.6/8.0
0
13
34
48
78
m-phase
63
300
3.1/2.4
0
-6
-13
-9
12
t-phase
63
900
3.1/2.4
0
-9
-18
-3 5
30
t-phase
63
300
2.0/1.0
0
-36
-88
-98
-92
o-phase
66
H ZO
300
bulk
0
24
48
58
95
80
9/4.5
0
-12
-28
-21
19
f-phase
11
300
9/4.5
0
-12
-28
-27
1
f-phase
11
900
9/4.5
0
-13
-30
-51
-43
t-phase
300
7.0/3.5
0
-26
-57
-59
-35
f,t-phase
-
11
2
16
300
5.5/3.0
0
-36
-78
-83
-61
t-phase
300
9.2/4.6
0
-14
-3 2
-31
-4
f-phase
16
300
14/7
0
-1
-4
-1
30
f-phase
16
300
19/9.5
0
5
9
15
47
m-phase
16
300
24/12
0
9
17
24
57
m-phase
16
37
100
106
72
m-phase
-63
t-phase
11
-84
t-phase
11
11
ZrO 2
300
bulk
0
24
34
80
9/4.5
0
-41
-95
300
9/4.5
0
-41
-96
900
9/4.5
0
-43
-99
-129
-137
t-phase
300
24/12
0
0
-15
-17
13
m-phase
300
36/18
0
8
1
1
33
3
62
m-phase
The results of the free energy model for the 9 nm film data series of Müller
11
in comparison
with bulk free energy is shown in FIG. 8. HfO2 is monoclinic for all temperatures. HZO is
ferroelectric for room temperature and below. ZrO2 is tetragonal for all temperatures. The
35
ferroelectricity in HZO arises in the model from the linear dependence of the surface energy on
the composition and the increase of the surface energy with crystal symmetry.
FIG. 8. Calculated Helmholtz free energy F for bulk (compare FIG. 3) and a 9 nm polycrystalline film for (a)
HfO2, (b) HZO, and (c) ZrO2 above 80 K. The colored area guides the eye to the difference in Helmholtz free
energy F of bulk (intersection with the ordinate) and the 9 nm thin film at 80 K.
The conclusion from the comparison of the Helmholtz free energy model with available data
is that the model describes the observations very well, although the involved energy differences
are in the order of a few meV/f.u. This is possible with a model, where the error from the total
energy contribution U and the error from the entropy contribution TS have been reduced to a
small amount.
As a consequence, the model achieves a predictive capability. FIG. 9 shows the modeled
Helmholtz free energy for the phases of HfO2, HZO, and ZrO2 relative to the phase with the
lowest Helmholtz free energy at these conditions as a function of film thickness with cylindrical
grains.
36
FIG. 9. Calculated free energy difference F relative to the phase with minimal Helmholtz free energy at 300 K
as a function of film thickness for (a) HfO 2, (b) HZO and (c) ZrO2. A window of stability for the f-phase arises for
HfO2 from 3 nm to 5 nm and in HZO from 8 nm and 16 nm thin film strength
The crossover of HfO2 from the high symmetry phases to the low symmetric m-phase
happens for film thickness below 5 nm. The model predicts ferroelectric HfO2 for a grain size
between 3 nm and 5 nm, for smaller grains the t-phase and c-phase. For HZO a similar
crossover exists but is shifted to larger grain size. The ferroelectric films exist in the thickness
regime between 8 nm and 16 nm. Thinner films or smaller grains occur in the t-phase. In
addition, our model suggests the existence of a ferroelectric phase in epitaxial HZO below a
critical film strength of about 5 nm. For ZrO2, the large surface energy difference of the tphase is dominant such that no f-phase is stable, although it is energetically close. The energy
disadvantage to the t-phase is small so that it can be overcome with an electric field (see Sec.
F). Despite being the second most stable phase in the bulk, the o-phase is never the most
stable phase for any grain volume.
F. Electric Enthalpy and Field D riven P hase Transform ation
37
The antiferroelectric behavior observed in doped HfO2
a field induced phase transformation
1
and ZrO2
11
has been interpreted as
19
. We examine the consistency of this statement with the
Helmholtz free energy model containing a contribution for a polarized crystal in an electric
field as described by Souza
69
. The Helmholtz free energy in a volume V0 with the field energy
contribution is
F,E = F - V0 DE= U – TS +  Ω - V0 DE
(7)
when D=D(E) = εrε0E+P is the electric displacement as a function of the macroscopic
electric field E which has the polarization P as a permanent contribution and εrε0E as an
induced contribution where εr is the dielectric tensor. The total energy is obtained from a
structural relaxation of the supercell including the field term. An induced polarization lowers
the energy while the field strain of the crystal increases U. A further effect is a modification of
the entropy from the field strain shifted phonon modes. However, we have neglected this
contribution.
We have calculated the Helmholtz free energy FE for ZrO2 bulk in the m-, t- and f-phase
and show the difference relative to the m-phase at zero E-field in FIG. 10 a-b. A field in ydirection orthogonal to the polarization direction does not lead to a phase transformation for
realizable fields. A field in the z-direction parallel to the polarization lowers the Helmholtz free
energy of the f-phase significantly by about 10 meV/f.u. for 1 MV/cm. For bulk ZrO 2 a field of
38
about 4 MV/cm would be required to give the f-phase a Helmholtz free energy below the mphase.
But for thin film ZrO2 the Helmholtz free energy of the t- and f-phase is already below the
m-phase. Here a much lower field of about 1 MV/cm is sufficient for the f-phase to obtain the
lowest energy. We can reproduce the temperature dependence of the field driven phase
transformation in 9nm thin films measured by Müller
11
by calculating the change of the
entropy contribution with temperature. In FIG. 10c we show the Helmholtz free energy of the
t- and f-phase for 80 K and 230 K. At 80 K a field strength of only 0.5 MV/cm is sufficient to
induce the phase transformation.
FIG. 10. Calculated Helmholtz free energy difference F,E for bulk ZrO2 relative to the m-phase at zero E-field
and T=300K, (a) for an E-field in y-direction, (b) for an E-field in z-direction and a polarization P in negative zdirection. (c) Shift of the energy crossover in 9 nm ZrO2 for the same E-field/Polarization configuration for a
temperature increase from T=80K to T=230K. For the t-phase the Helmholtz free energy difference to the mphase F,E is dependent on temperature, while for the f-phase it is nearly independent on temperature (see FIG.
3).
The necessary E-field for the field driven phase transformation from the t- to the f-phase was
so far determined from the Helmholtz free energy difference F,E(f-phase)- F,E(t-phase) in an
equilibrium state. A further requirement for a phase transformation is a sufficiently low energy
39
barrier to the new Helmholtz free energy minimum. As a good approximation to the transition
barrier34 we have calculated the maximum of the transition path from the t- to the f-phase.
We obtain a barrier of only 30 meV/f.u. for the chosen orientation in TABLE I. Such a small
barrier should be overcome by thermal energy alone, hence removal of the external field relaxes
the structure back to the t-phase.
Similar values have been recently found in the literature (30 meV/f.u. in Huan
meV/f.u. from Reyes-Lillo
47
49
and 30
). The low barrier is related to a soft phonon mode interpolating
between t- and f-phase. Reyes-Lillo et al. found a total energy difference of only 1 meV
between f- and t-phase and calculated a critical field strength of 1.2 MV/cm necessary to
overcome the barrier of 30 meV/f.u. They do not explain why the f-phase relaxes back into the
t-phase, after the field strength is reduced to zero.
To show the consistency of the Helmholtz free energy calculation we have extracted the
dielectric constants ε ri, i=x,y,z (diagonal elements of the dielectric tensor) and the polarization
P from a second order fit to the computed values of the field energy V0 DE= V0(εrε0E+P)E.
TABLE VII shows the comparison of these values εE-field for m-, f- and t-HfO2 with values εionic
from a conventional linear response calculation to an ionic perturbation and εtotal including the
electronic contribution. The results only show a small deviation between εE-field and εionic
indicating a slightly nonlinear behavior.
40
TABLE VII. Diagonal elements of the dielectric tensor and averaged dielectric constants calculated from the
Helmholtz free energy and from a linear response calculation.
H fO2
H ZO
ZrO2
εE- εionic εtotal
εionic
εtotal
εionic
εtotal
field
m -phase P 14
x
y
z
Σ/3
20.0 19.8
18.0 18.1
15.0 15.0
17.7 17.7
24.7 21.5
22.9 19.6
19.5 15.6
22.4 18.9
26.7
24.8
20.5
24.1
23.3
21.3
16.3
20.3
29.0
27.0
21.5
25.8
23.7
21.6
16.8
20.7
29.5
27.3
22.1
26.3
28.4
22.1
26.1
25.5
34.4
27.8
31.7
31.3
o-phase P 61
x
y
z
Σ/3
-
20.3
18.0
15.5
17.9
25.3
22.8
20.1
22.7
-
f-phase P 29
x
y
z
Σ/3
24.0
19.0
26.0
23.0
23.8
19.3
23.1
22.1
28.8 26.0
24.2 20.7
28.0 24.6
27.0 23.8
31.5
25.9
29.9
29.1
t-phase P 137
x,z 55.0 51.8 56.9 54.3 60.4 55.4 61.5
y 20.0 19.2 24.0 19.9 25.3 19.9 25.3
Σ/3 43.3 40.9 45.9 42.8 48.7 43.6 49.4
c-phase P 225
x,y,z
-
Σ/3
-
30.9 36.0
30.9
36.0
-
41.5 47.6
-
41.5 47.6
The comparison of the extracted polarization PE-field for f-HfO2 with values PBerry from a
conventional Berry phase calculation is shown in TABLE VIII. The values match excellent.
Furthermore, the table shows the dependence of the polarization P of Hf1-χZrχO2 on the
stoichiometry χ. The polarization increases with Zr content. This can be understood by the
increasing values of the Born charges obtained from the linear response calculation. TABLE
VIII shows only the diagonal values of the Born charge tensor. A simplified estimation based
on equation (1) including the diagonal values and a perturbation in z-direction around the
41
centrosymmetric configuration leads to a good estimation and explains the stoichiometric trend
of the polarization.
42
TABLE VIII. Polarization
H fO2
P E-
H ZO
ZrO2
P B erry
P B erry
P B erry
0.504
0.541
0.579
field
z
0.505
ε E-
z
z
z
field
Z
Z
Z
Zr
O1
O2
20.0 0.003 5.08 0.002 5.23 0.002 5.85
18.0 -0.138 -2.53 -0.135 -2.54 -0.133 -2.58
15.0 0.008 2.52 0.008 2.58 0.006 2.60
eq.
(1)
17.7
0.67
0.69
0.72
IV . C ON C LU SION
In this paper we have investigated the origin of the ferroelectricity in Hf1-χZrχO2 for HZO
with χ = 0.5, and the origin of antiferroelectricity for ZrO2 with DFT calculations, and a
phenomenological model for the surface energy contribution. From total energy calculations we
found that the bulk ferroelectric phase Pca21 is not stable. The computations for several
polymorphs were compared to a number of literature results to estimate the theoretical
uncertainty and find the most reliable method. We decided for the LDA functional and normconserving TM pseudopotentials. The temperature dependent Helmholtz free energy was
calculated accordingly. As a first possible mechanism to stabilize the ferroelectric phase we
investigated the film strain. Although we could reproduce all relevant stress and strain
dependent phenomena in HfO2 and ZrO2 with our model, we only found a small window of
rather high compressive film strain to allow a stabilization of the ferroelectric phase. Since the
thus stabilized structure has a polarization in the strain plane, which is unobservable, we found
43
the stress/strain mechanism unlikely to be the major cause of the stabilization. As a second
possible mechanism we investigated the surface energy contribution to the Helmholtz free
energy. Since the surface energy of ferroelectric grains has not been measured or calculated so
far we build a phenomenological model for all polymorphs and all stoichiometry values by
interpolating between existing values using decreasing values for increasing crystal symmetry.
With the Gibbs- / Helmholtz free energy model containing an ab initio computed part for total
energy, entropy, stress or strain, and a phenomenological part for the surface energy we could
reproduce the observed phases for nanosized thin films and grains and found stable
ferroelectric HZO in a size window around 10 nm at room temperature in absence of strain.
Based on the model we furthermore predict a similar ferroelectric window in HfO 2 around 4
nm and below 5 nm in epitaxial HZO. For ZrO2 no stable ferroelectric nanosized grains exist.
After calculating the field dependent contribution to the Helmholtz free energy self-consistently
we found an electric field of about 1 MV/cm sufficient to give ferroelectric grains of 9 nm at
room temperature the lowest Helmholtz free energy. Furthermore, reducing the temperature
favors the stability and decreases the crossover field in accordance with experimental data.
A C K N OW LED G M EN TS
The author wants to thank U. Schroeder and T. Schenk from NamLab and U. Böttger and S.
Starschich from RWTH Aachen for discussions. The German Research Foundation (Deutsche
Forschungsgemeinschaft) is acknowledged for funding this research in the frame of the project
“Inferox” (project no. MI 1247/11-1). The authors gratefully acknowledge the Gauss Centre for
44
Supercomputing e.V. (www.gauss-centre.eu) for funding this project by providing computing
time on the GCS Supercomputer SuperMUC at Leibniz Supercomputing Centre (LRZ,
www.lrz.de).
45
1. T. S. Böscke, J. Müller, D. Bräuhaus, U. Schröder, and U. Böttger, “Ferroelectricity in
hafnium oxide thin films,” Appl. Phys. Lett. 99 (10), 102903 (2011).
2. J. Müller, T. S. Böscke, D. Bräuhaus, U. Schröder, U. Böttger, J. Sundqvist, P. Kücher, T.
Mikolajick, and L. Frey, “Ferroelectric Zr0.5Hf0.5O2 thin films for nonvolatile memory
applications,” Appl. Phys. Lett. 99 (11), 112901 (2011).
3. J. Wang, H. P. Li, and R. Stevens, “Hafnia and hafnia-toughened ceramics,” J Mater Sci 27
(20), 5397–5430 (1992).
4. E. H. Kisi and C. J. Howard, “Crystal Structures of Zirconia Phases and their InterRelation,” KEM 153-154, 1–36 (1998).
5. E. H. Kisi, C. J. Howard, and R. J. Hill, “Crystal Structure of Orthorhombic Zirconia in
Partially Stabilized Zirconia,” J American Ceramic Society 72 (9), 1757–1760 (1989).
6. J. Lowther, J. Dewhurst, J. Leger, and J. Haines, “Relative stability of ZrO2 and HfO2
structural phases,” Phys. Rev. B 60 (21), 14485–14488 (1999).
7. J. H. Choi, Y. Mao, and J. P. Chang, “Development of hafnium based high-k materials—A
review,” Materials Science and Engineering: R: Reports 72 (6), 97–136 (2011).
8. S. Mueller, J. Mueller, A. Singh, S. Riedel, J. Sundqvist, U. Schroeder, and T. Mikolajick,
“Incipient Ferroelectricity in Al-Doped HfO2 Thin Films,” Adv. Funct. Mater. 22 (11),
2412–2417 (2012).
9. T. Olsen, U. Schröder, S. Müller, A. Krause, D. Martin, A. Singh, J. Müller, M. Geidel,
and T. Mikolajick, “Co-sputtering yttrium into hafnium oxide thin films to produce
ferroelectric properties,” Appl. Phys. Lett. 101 (8), 082905 (2012).
10. Min Hyuk Park, Han Joon Kim, Yu Jin Kim, Young Hwan Lee, Taehwan Moon, Keum
Do, “Ferroelectricity and Antiferroelectricity of Doped Thin HfO2 -based Films,” Advanced
Materials (2015).
11. J. Müller, T. S. Böscke, U. Schröder, S. Mueller, D. Bräuhaus, U. Böttger, L. Frey, and T.
Mikolajick, “Ferroelectricity in Simple Binary ZrO 2 and HfO 2,” Nano Lett. 12 (8), 4318–
4323 (2012).
12. J. Muller, T. S. Boscke, U. Schroder, R. Hoffmann, T. Mikolajick, and L. Frey,
“Nanosecond Polarization Switching and Long Retention in a Novel MFIS-FET Based on
Ferroelectric $\hbox{HfO}_{2}$,” IEEE Electron Device Lett. 33 (2), 185–187 (2012).
13. J. Müller, U. Schröder, T. S. Böscke, I. Müller, U. Böttger, L. Wilde, J. Sundqvist, M.
Lemberger, P. Kücher, T. Mikolajick, and L. Frey, “Ferroelectricity in yttrium-doped
hafnium oxide,” J. Appl. Phys. 110 (11), 114113 (2011).
14. D. Zhou, J. Müller, J. Xu, S. Knebel, D. Bräuhaus, and U. Schröder, “Insights into
electrical characteristics of silicon doped hafnium oxide ferroelectric thin films,” Appl. Phys.
Lett. 100 (8), 082905 (2012).
15. J. Robertson, “High dielectric constant gate oxides for metal oxide Si transistors,” Rep.
Prog. Phys. 69 (2), 327–396 (2006).
46
16. M. Hyuk Park, H. Joon Kim, Y. Jin Kim, T. Moon, and C. Seong Hwang, “The effects of
crystallographic orientation and strain of thin Hf0.5Zr0.5O2 film on its ferroelectricity,”
Appl. Phys. Lett. 104 (7), 072901 (2014).
17. R. C. Garvie, “The Occurrence of Metastable Tetragonal Zirconia as a Crystallite Size
Effect,” J. Phys. Chem. 69 (4), 1238–1243 (1965).
18. E. H. Kisi, “Influence of Hydrostatic Pressure on the t→o Transformation in Mg-PSZ
Studied by In Situ Neutron Diffraction,” Journal of the American Ceramic Society 81 (3),
741–745 (1998).
19. T. S. Böscke, St. Teichert, D. Bräuhaus, J. Müller, U. Schröder, U. Böttger, and T.
Mikolajick, “Phase transitions in ferroelectric silicon doped hafnium oxide,” Appl. Phys.
Lett. 99 (11), 112904 (2011).
20. M. W. Pitcher, S. V. Ushakov, A. Navrotsky, B. F. Woodfield, G. Li, J. Boerio-Goates,
and B. M. Tissue, “Energy Crossovers in Nanocrystalline Zirconia,” Journal of the
American Ceramic Society 88 (1), 160–167 (2005).
21. S. V. Ushakov, A. Navrotsky, Y. Yang, S. Stemmer, K. Kukli, M. Ritala, M. A. Leskelä, P.
Fejes, A. Demkov, C. Wang, B.-Y. Nguyen, D. Triyoso, and P. Tobin, “Crystallization in
hafnia- and zirconia-based systems,” phys. stat. sol. (b) 241 (10), 2268–2278 (2004).
22. A. Christensen and E. Carter, “First-principles study of the surfaces of zirconia,” Phys.
Rev. B 58 (12), 8050–8064 (1998).
23. W. Zhou, S. V. Ushakov, T. Wang, J. G. Ekerdt, A. A. Demkov, and A. Navrotsky,
“Hafnia: Energetics of thin films and nanoparticles,” J. Appl. Phys. 107 (12), 123514
(2010).
24. M. Hyuk Park, H. Joon Kim, Y. Jin Kim, W. Lee, T. Moon, and C. Seong Hwang,
“Evolution of phases and ferroelectric properties of thin Hf0.5Zr0.5O2 films according to
the thickness and annealing temperature,” Appl. Phys. Lett. 102 (24), 242905 (2013).
25. X. Gonze, “A brief introduction to the ABINIT software package,” Zeitschrift für
Kristallographie 220 (5/6/2005) (2005).
26. X. Gonze, B. Amadon, P.-M. Anglade, J.-M. Beuken, F. Bottin, P. Boulanger, F.
Bruneval, D. Caliste, R. Caracas, M. Côté, T. Deutsch, L. Genovese, P. Ghosez, M.
Giantomassi, S. Goedecker, D. R. Hamann, P. Hermet, F. Jollet, G. Jomard, S. Leroux, M.
Mancini, S. Mazevet, M. Oliveira, G. Onida, Y. Pouillon, T. Rangel, G.-M. Rignanese, D.
Sangalli, R. Shaltaf, M. Torrent, M. J. Verstraete, G. Zerah, and J. W. Zwanziger,
“ABINIT: First-principles approach to material and nanosystem properties,” Computer
Physics Communications 180 (12), 2582–2615 (2009).
27. M. Fuchs and M. Scheffler, “Ab initio pseudopotentials for electronic structure calculations
of poly-atomic systems using density-functional theory,” Computer Physics
Communications 119 (1), 67–98 (1999).
28. D. Porezag, M. Pederson, and A. Liu, “Importance of nonlinear core corrections for densityfunctional based pseudopotential calculations,” Phys. Rev. B 60 (20), 14132–14139 (1999).
47
29. H. J. Monkhorst and J. D. Pack, “Special points for Brillouin-zone integrations,” Phys. Rev.
B 13 (12), 5188–5192 (1976).
30. X. Gonze, “Dynamical matrices, Born effective charges, dielectric permittivity tensors, and
interatomic force constants from density-functional perturbation theory,” Phys. Rev. B 55
(16), 10355–10368 (1997).
31. A. Rappe, K. Rabe, E. Kaxiras, and J. Joannopoulos, “Optimized pseudopotentials,” Phys.
Rev. B 41 (2), 1227–1230 (1990).
32. See supplementary material at [ULR] for additional data sets concerning structural data,
IR spectra, stress and strain, surface energy, and polarization.
33. W. Zheng, K. H. Bowen, J. Li, I. Da̧ bkowska, and M. Gutowski, “Electronic Structure
Differences in ZrO 2 vs HfO 2†,” J. Phys. Chem. A 109 (50), 11521–11525 (2005).
34. X. Luo, W. Zhou, S. Ushakov, A. Navrotsky, and A. Demkov, “Monoclinic to tetragonal
transformations in hafnia and zirconia: A combined calorimetric and density functional
study,” Phys. Rev. B 80 (13) (2009).
35. H. Arashi, “Pressure-Induced Phase Transformation of HfO2,” J American Ceramic Society
75 (4), 844–847 (1992).
36. Y. Al-Khatatbeh, Lee, Kanani K. M., and B. Kiefer, “Phase diagram up to 105 GPa and
mechanical strength of HfO_{2},” Phys. Rev. B 82 (14) (2010).
37. J. Leger, P. Tomaszewski, A. Atouf, and A. Pereira, “Pressure-induced structural phase
transitions in zirconia under high pressure,” Phys. Rev. B 47 (21), 14075–14083 (1993).
38. Y. Al-Khatatbeh, Lee, Kanani K. M., and B. Kiefer, “Phase relations and hardness trends
of ZrO_{2} phases at high pressure,” Phys. Rev. B 81 (21) (2010).
39. K. Momma and F. Izumi, “VESTA 3 for three-dimensional visualization of crystal,
volumetric and morphology data,” J Appl Crystallogr 44 (6), 1272–1276 (2011).
40. H. T. Stokes and D. M. Hatch, “FINDSYM. Program for identifying the space-group
symmetry of a crystal,” J Appl Crystallogr 38 (1), 237–238 (2005).
41. A. G. Jackson, Handbook of Crystallography. For Electron Microscopists and Others
(Springer New York, New York, NY, 1991).
42. K. Negita and H. Takao, “Condensations of phonons at the tetragonal to monoclinic phase
transition in ZrO2,” Journal of Physics and Chemistry of Solids 50 (3), 325–331 (1989).
43. D. Simeone, G. Baldinozzi, D. Gosset, M. Dutheil, A. Bulou, and T. Hansen, “Monoclinic
to tetragonal semireconstructive phase transition of zirconia,” Phys. Rev. B 67 (6) (2003).
44. G. Csonka, J. Perdew, A. Ruzsinszky, P. Philipsen, S. Lebègue, J. Paier, O. Vydrov, and J.
Ángyán, “Assessing the performance of recent density functionals for bulk solids,” Phys.
Rev. B 79 (15) (2009).
45. R. RUH and CORFIELD, PETER W. R., “Crystal Structure of Monoclinic Hafnia and
Comparison with Monoclinic Zirconia,” J American Ceramic Society 53 (3), 126–129
(1970).
46. Q. Zeng, A. R. Oganov, A. O. Lyakhov, C. Xie, X. Zhang, J. Zhang, Q. Zhu, B. Wei, I.
Grigorenko, L. Zhang, and L. Cheng, “Evolutionary search for new high- k dielectric
48
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
materials: methodology and applications to hafnia-based oxides,” Acta Crystallogr C Struct
Chem 70 (2), 76–84 (2014).
S. E. Reyes-Lillo, K. F. Garrity, and K. M. Rabe, “Antiferroelectricity in thin-film ZrO2
from first principles,” Phys. Rev. B 90 (14) (2014).
J. Jaffe, R. Bachorz, and M. Gutowski, “Low-temperature polymorphs of ZrO2 and HfO2:
A density-functional theory study,” Phys. Rev. B 72 (14) (2005).
T. D. Huan, V. Sharma, G. A. Rossetti, and R. Ramprasad, “Pathways towards
ferroelectricity in hafnia,” Phys. Rev. B 90 (6) (2014).
I. MacLaren, T. Ras, M. MacKenzie, A. J. Craven, D. W. McComb, and S. de Gendt,
“Texture, Twinning, and Metastable “Tetragonal” Phase in Ultrathin Films of HfO[sub 2]
on a Si Substrate,” J. Electrochem. Soc. 156 (8), G103 (2009).
R. Suyama, H. Takubo, and S. Kume, “Synthesis of Hf1-x, Zrx02 (O,” J American Ceramic
Society 68 (9), C-237 (1985).
I. A. EL-SHANSHOURY, V. A. RUDENKO, and I. A. IBRAHIM, “Polymorphic Behavior
of Thin Evaporated Films of Zirconium and Hafnium Oxides,” Journal of the American
Ceramic Society 53 (5), 264–268 (1970).
G. Fadda, G. Zanzotto, and L. Colombo, “First-principles study of the effect of pressure on
the five zirconia polymorphs. I. Structural, vibrational, and thermoelastic properties,” Phys.
Rev. B 82 (6) (2010).
O. Ohtaka, T. YAMANAKA, S. Kume, N. HARA, H. ASANO, and F. Izumi, “Structural
analysis of orthorhombic ZrO2 by high resolution neutron powder diffraction,” Proceedings
of the Japan Academy. Ser. B: Physical and Biological Sciences 66 (10), 193–196 (1990).
F. Namavar, G. Wang, C. L. Cheung, R. F. Sabirianov, X. C. Zeng, W. N. Mei, J. Bai, J.
R. Brewer, H. Haider, and K. L. Garvin, “Thermal stability of nanostructurally stabilized
zirconium oxide,” Nanotechnology 18 (41), 415702 (2007).
X. Zhao and D. Vanderbilt, “First-principles study of structural, vibrational, and lattice
dielectric properties of hafnium oxide,” Phys. Rev. B 65 (23) (2002).
C. J. Howard, E. H. Kisi, and O. Ohtaka, “Crystal Structures of Two Orthorhombic
Zirconias,” J American Ceramic Society 74 (9), 2321–2323 (1991).
C. Pecharromán, M. Ocaña, and C. J. Serna, “Optical constants of tetragonal and cubic
zirconias in the infrared,” J. Appl. Phys. 80 (6), 3479 (1996).
D. A. Neumayer and E. Cartier, “Materials characterization of ZrO[sub 2]–SiO[sub 2] and
HfO[sub 2]–SiO[sub 2] binary oxides deposited by chemical solution deposition,” J. Appl.
Phys. 90 (4), 1801 (2001).
O. Ohtaka, T. YAMANAKA, S. Kume, N. HARA, H. ASANO, and F. Izumi, “Structural
Analysis of Orthorhombic Hafnia by Neutron Powder Diffraction,” J American Ceramic
Society 78 (1), 233–237 (1995).
O. Diéguez, K. Rabe, and D. Vanderbilt, “First-principles study of epitaxial strain in
perovskites,” Phys. Rev. B 72 (14) (2005).
49
62. M. Yashima and S. Tsunekawa, “Structures and the oxygen deficiency of tetragonal and
monoclinic zirconium oxide nanoparticles,” Acta Crystallogr B Struct Sci 62 (1), 161–164
(2006).
63. M. Shandalov and P. C. McIntyre, “Size-dependent polymorphism in HfO[sub 2] nanotubes
and nanoscale thin films,” J. Appl. Phys. 106 (8), 084322 (2009).
64. J. Chevalier, L. Gremillard, A. V. Virkar, and D. R. Clarke, “The Tetragonal-Monoclinic
Transformation in Zirconia: Lessons Learned and Future Trends,” Journal of the American
Ceramic Society 92 (9), 1901–1920 (2009).
65. A. V. Radha, O. Bomati-Miguel, S. V. Ushakov, A. Navrotsky, and P. Tartaj, “Surface
Enthalpy, Enthalpy of Water Adsorption, and Phase Stability in Nanocrystalline
Monoclinic Zirconia,” Journal of the American Ceramic Society 92 (1), 133–140 (2009).
66. F. Bohra, B. Jiang, and J.-M. Zuo, “Textured crystallization of ultrathin hafnium oxide
films on silicon substrate,” Appl. Phys. Lett. 90 (16), 161917 (2007).
67. W. G. Wolfer, “Elastic properties of surfaces on nanoparticles,” Acta Materialia 59 (20),
7736–7743 (2011).
68. B. Santara, P. K. Giri, K. Imakita, and M. Fujii, “Microscopic origin of lattice contraction
and expansion in undoped rutile TiO 2 nanostructures,” J. Phys. D: Appl. Phys. 47 (21),
215302 (2014).
69. I. Souza, J. Íñiguez, and D. Vanderbilt, “First-Principles Approach to Insulators in Finite
Electric Fields,” Phys. Rev. Lett. 89 (11) (2002).
50